A Reinforcement Learning-Adapter Renewable Energy Forecasting Framework for Two-Stage Distribution Network Scheduling

View at publisher Summarise with VIRAT
Authors: Chenghan Li, Yinliang Xu, Hongbin Sun
Cited by: 0
References: 37

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

With deepening renewable penetration, distribution-network operators must rely on high-quality forecasts to schedule distributed energy resources. However, most existing Prediction– then–Optimization (PTO) approaches rely solely on forecasted values as inputs, while overlooking the significant impact of the coupling between actual and forecasted values on operational costs. This limitation hinders the practicality and deployment of many decision-oriented methods in power systems. To address this issue, this paper investigates the two-stage scheduling problem of distribution networks and formulates it as a Coupled Prediction–then–Optimization (CPTO) problem. Then, we propose a RL-driven decision-focused framework that integrates a Generative Pretrained Transformer (GPT)-based forecasting module with an improved Actor–Critic Reinforcement Learning (RL) adapter. Specifically, a GPT-based forecasting model with task-adaptive loss is proposed to capture the temporal characteristics of renewable energy outputs. On this basis, an RL adapter is designed to learn a decision-focused policy that dynamically refines the base forecasts, thereby effectively reducing the actual operating cost of the distribution network. Case studies on the IEEE 33-bus, IEEE 141-bus and IEEE 570-bus systems demonstrate that, compared with best baselines, the proposed framework reduces operational costs by 2.20% , 2.28% and 1.80% respectively, thereby validating its effectiveness and superiority in terms of both economic efficiency and operational reliability.

Fields of Study & Subjects

Optimal Power Flow Distribution Electric Power System Optimization Smart Grid and Power Systems

Publication details

Journal / source
IEEE Transactions on Smart Grid
Volume
17
Issue
5
Pages
3776–3788
Publication date
25 March 2026
Language
EN